How to get named when a buyer asks ChatGPT for suppliers

A buyer at a packaging machine maker near Stuttgart needs 40 aluminium prototype housings, machined on five axes, in three weeks. He asks ChatGPT: “Which companies in Bulgaria do 5-axis CNC machining of aluminium prototypes and hold ISO 9001?” Seconds later he has four names, each with a line of facts and a link to its source.

He opens two links and sends one RFQ. If your company was not among the four, you lost nothing you can see: you were simply never on the list. 6sense found that 95% of the time, the winning vendor was already on the buyer's day-one shortlist (2025). An assistant's answer is now one of the places where that shortlist gets written.

Here is my AI search playbook as six steps you can start this week. For the whole research path, from search to datasheet to RFQ, see how industrial buyers choose suppliers.

How an assistant decides whom to name

For a question like this, ChatGPT, Gemini and Perplexity usually search the web and name the companies whose pages answer it, linking each source. Google says AI Overviews and AI Mode may use “query fan-out”, issuing multiple related searches across subtopics and data sources to build one response. So the buyer's question becomes several searches: 5-axis, aluminium, prototypes, Bulgaria, ISO 9001.

A page that states each fact in words matches each of those searches. “Precision engineering solutions for demanding clients” matches none. A typical answer looks like this.

What you do not need is just as clear. Google says there are no additional requirements or special optimisations for AI Overviews and AI Mode, and no need for new machine-readable files, AI text files or markup. The page must be indexed and eligible to show with a snippet.

At ITCE, a B2B IT training and ServiceNow consulting company in Sofia Tech Park, the early numbers already justify the work.

The volume is small, 77 visits against 2,512 organic and AI sessions, but the quality is not. ITCE sells training, not machines, yet its buyers ask what engineers ask: who can do this, where, and with which certification.

Ask the questions your buyers ask

Measure first, or you will have nothing to compare against. Write ten to twenty questions the way a buyer would type them and ask each in ChatGPT, Gemini and Perplexity. Record who is named, in what order, and which page is cited.

Mix supplier lists (“best suppliers of laser-cut stainless parts in Southeast Europe”), spec questions (“who can machine titanium to ±0.01 mm in Bulgaria”) and comparisons (“5-axis or 3-axis machining for 50 aluminium parts”), some in German or Romanian if you export. Take the wording from Search Console and your customers' RFQs; finding the searches that come before a quote request shows how.

Start a new chat each time and screenshot every answer. Common mistake: asking “What does [your company] do?”. That tests whether the assistant knows you, not whether it recommends you to a stranger.

State what you make, and put the specs in text

Most reasons for being left out sit on your own site. Only 38.7% of Bulgarian companies have a website that describes their products or lists prices, against 66.4% across the EU (Eurostat, 2025).

Say it in one paragraph, on the homepage and the About page, in the third person, so it still makes sense when quoted alone: “Company X is a CNC machining subcontractor in Plovdiv. It machines aluminium, steel and titanium parts up to 800 × 500 × 400 mm on 5-axis machines, in batches of 1 to 500, for packaging machinery and medical device makers in Germany, Austria and Bulgaria. It is certified to ISO 9001:2015.”

Common mistake: certificates shown only as logos in the footer, and “a dynamic team with 25 years of experience”. Neither gives a buyer anything to check.

Then put the specs on the page as text: dimensions, materials, tolerances, lead times and standards, in a table on each capability page. Google's AI features guidance lists “making sure that important content is available in textual form” as a best practice. Common mistake: a “Download catalogue” button as the only place the numbers exist. Capability pages that win RFQs covers the page row by row.

Then check that assistants can reach you at all. OpenAI says sites that block its search crawler, OAI-SearchBot, will not be shown in ChatGPT search answers; blocking GPTBot, its training crawler, is a separate setting. Google advises allowing crawling in robots.txt and by any CDN or hosting infrastructure.

Write the comparison pages your competitors avoid

Buyers ask assistants to compare, and Google says AI Mode is particularly helpful where “complex comparisons are needed”. An assistant can only cite a comparison someone has written. Search your own field: often no supplier has written one.

Write “how to choose” pages (choosing a gearbox repair shop or a sheet metal subcontractor) and honest “X or Y” pages (laser or waterjet for 10 mm stainless, 5-axis or 3+2 for small batches, cobot or industrial robot for palletising).

Fair means saying when the answer is not you. A line such as “Above 2,000 pieces a year, a high-volume turning shop will usually be cheaper” costs you no orders and tells every reader, human or assistant, exactly who you are right for.

End each page with which of your services fits which case, linked to its capability page. Common mistake: a “comparison” you win on every row. Engineers stop trusting it by the second row.

Be listed where assistants look

Look again at the links under your recorded answers. Many are not company sites but directories, association lists, trade articles and distributor pages: places where an assistant finds you, or a competitor in your place.

Use the same facts everywhere: name, products, certificates and countries should match your paragraph from step two. Common mistake: hundreds of links from low-quality directories. Ten correct entries on sites buyers read are worth more than two hundred on sites nobody visits.

Score yourself once a month

On the first working day of each month, ask the same questions in the same three assistants. Your score is the share of answers that name you. Log the competitors named in each answer in a second tab. Common mistake: changing the questions every month, which makes the scores impossible to compare.

Watch the last column too. If a directory or an old PDF is cited for you, your own page is the next thing to fix.

Then check what the visits are worth. ChatGPT adds utm_source=chatgpt.com to its links, so those visits show up by source in Google Analytics. Since 31 August 2026, Search Console's generative AI reports, now open to every site, show how often your pages appear in AI Overviews and AI Mode, by page, country and device. Count leads, not visits: tracking every RFQ back to the search shows how.

At ITCE, Semrush counted 48 itce.com pages cited by AI engines on 17 September 2026. None of the work was AI-specific: I merged the ServiceNow pages into one hub, rewrote 849 titles and 531 descriptions, rewrote 652 vague links and added 214 internal links. I cannot prove which change moved the AI numbers, but it is what Google describes: clear pages that can be found and read.

One honest note: forms from organic and AI search rose from 303 to 423 while confirmed leads dipped from 87 to 76, so a form audit and a monthly check against the CRM are the next step there.

What to do this week

1. Ask ten buyer questions about your best-margin product line in all three assistants; fill in the sheet.
2. Check robots.txt for OAI-SearchBot and Googlebot; ask your host about the firewall and CDN.
3. Put your one-paragraph statement on the homepage and the About page.
4. Move one capability page's key specs out of the PDF into a text table.
5. Correct your entries on five directories, association lists or distributor sites.

Then set a reminder for the first working day of next month. If you would rather run the first test with me, book a call below.